Error detection and concealment using predicted signal values

Abstract

The system transmits blocks of N signal values preceded by block code words which are used at the receiver to detect errors. In the transmitter, the digital value of the intelligence signal being transmitted is sent to a predictor, a comparator, and an N-word delay line. The predictor estimates the next signal value, which is suitably delayed so that the current value and its estimate may be compared. The comparator subtracts each estimated value from the corresponding actual value, temporarily storing the difference, called a "delta value," while the entire block of N words is processed. A "Maximum Delta Encoder" encodes (with appropriate error correction) the absolute value of the maximum delta value for all N words. The resulting block code word is multiplexed into the data channel ahead of the group of N words to which it corresponds. An "N Sample Delay" is used to delay the N words so that the block code word will precede it. In the receiver, the block code word is demultiplexed and error corrected to recover the absolute value of the maximum delta value for the group of N words which follow. As each of the N words is received, it is subtracted from a predicted value generated by a predictor circuit identical to the one found in the transmitter. If the difference is greater than the maximum delta value for that block, the received signal value represents an error. The error is then concealed by replacing it with the predicted value itself. If the difference is less than or equal to the maximum delta value, the received signal is accepted as correct.

I claim: 
     
       1. A method of enabling the detection of errors introduced in an informal signal during transmission of the informal signal through a communication channel to a receiver, said method comprising the steps of: predicting from the untransmitted information signal a plurality of actual values of the information signal and generating therefrom a plurality of predicted values;   determining the difference between each of the plurality of actual values and the corresponding predicted value;   developing an error detection signal based on a comparison of the determined differences between each of the plurality of actual values and the corresponding predicted value;   transmitting the plurality of actual values through the channel to the receiver; and   transmitting the error detection signal to the receiver.   
     
     
       2. The method of claim 1 wherein the information signal is a correlated signal. 
     
     
       3. A method of detecting errors introduced into an information signal during passage of the information signal through a communication channel to a receiver, said method comprising the steps of: predicting from the untransmitted information signal a plurality of actual values of the information signal;   determining the difference between each of the plurality of actual values and the corresponding prediction;   developing an error detection signal from the determined differences between each of the plurality of actual values and the corresponding prediction;   transmitting the plurality of actual values through the channel to the receiver;   transmitting the error detection signal to the receiver;   receiving the transmitted plurality of actual values and the transmitted error detection signal at the receiver;   predicting from the received actual values each of the plurality of actual values of the received information signal;   determining the difference between each of the plurality of received actual values and the corresponding prediction; and   comparing the received error detection signal with determined differences between each of the plurality of received actual values in the corresponding prediction to detect errors.   
     
     
       4. The method of claim 3 wherein the information signal is a correlated signal. 
     
     
       5. A method of enabling the detection of errors introduced in an information signal during transmission of the information signal through a communication channel to a receiver, said method comprising the steps of; periodically sampling the information signal;   calculating from the untransmitted samples a predicted value corresponding to each sample;   determining the difference between each sample and the corresponding predicted value;   developing an error detection signal based on a comparison of the determined differences between each sample and the corresponding predicted value;   transmitting the samples through the channel to the receiver; and   transmitting the error detection signal to the receiver.   
     
     
       6. The method of claim 5 wherein the information signal has known statistical properties and said step of calculating from the untransmitted samples a predicted value corresponding to each sample comprises calculating the predicted value based solely on the value of one or more previous untransmitted samples and the statistical properties of the information signal. 
     
     
       7. A method of detecting errors introduced in an information signal during passage of the information signal through a communication channel to a receiver, said method comprising the steps of: periodically sampling the information signal;   calculating from the untransmitted samples a predicted value corresponding to each sample;   determining the difference between each sample and the corresponding predicted value;   developing an error detection signal from the determined differences between each sample and the corresponding predicted value;   transmitting the samples through the channel to the receiver;   transmitting the error detection signal to the receiver;   receiving the transmitted samples and the transmitted error detection signal at the receiver;   calculating from the received samples a predicted value corresponding to each of the samples;   determining the difference between each received sample and the corresponding predicted value; and   comparing the received error detection signal with the determined differences between each received sample and the corresponding predicted value to detect errors.   
     
     
       8. The method of claim 7 wherein: the information signal has known statistical properties;   said step of calculating from the untransmitted samples a predicted value comprises calculating the predicted value based solely on the value of one or more previous untransmitted samples and the statistical properties of the information signal; and   said step of calculating from the received samples a predicted value comprises calculating the predicted value based solely on the value of one or more previous received samples and the statistical properties of the information signal.   
     
     
       9. A method of detecting and concealing errors introduced into an information signal during passage of the information signal through a communication channel to a receiver, said method comprising the steps of: receiving a plurality of actual values of the information signal;   predicting from the received actual values each of the plurality of actual values of the received information signal;   determining the difference between each of the plurality of received actual values and the corresponding prediction;   receiving a maximum difference value;   comparing the received maximum difference value with the determined differences between each of the plurality of received actual values and the corresponding prediciton in order to detect errors; and   substituting, for any received actual value detected to be in error, the corresponding prediction.   
     
     
       10. The method of claim 9 further comprising the inital steps of: predicting from the untransmitted information signal a plurality of actual values of the information signal;   determining the difference between each of the plurality of actual values of the untransmitted information signal and the corresponding prediction;   evaluating the maximum value of the determined differences between each of the plurality of actual values of the untransmitted information signal and the corresponding prediction over a predetermined period of time; and   transmitting the evaluated maximum value to the receiver as the maximum difference value.   
     
     
       11. A method of detecting and concealing errors introduced into an information signal during passage of the information signal through a communication channel to a receiver, said method comprising the steps of: receiving a plurality of samples of the information signal;   calculating from the received samples a predicted value corresponding to each of the samples;   determining the difference between each received sample and the corresponding predicted value;   receiving a maximum difference value;   comparing the received maximum difference value with the determined differences between each received sample and the corresponding predicted value in order to detect errors; and   substituting, for any received sample detected to be in error, the corresponding predicted value.   
     
     
       12. The method of claim 11 further comprising the initial steps of: calculating from the untransmitted information signal a predicted value corresponding to each of a plurality of samples of the information signal;   determining the difference between each sample of the untransmitted information signal and the corresponding predicted value;   evaluating the maximum value of the determined differences between each sample of the untransmitted information signal and the corresponding predicted value for a predtermined number of samples; and   transmitting the evaluated maximum value to the receiver as the maximum difference value.   
     
     
       13. A method of detecting errors in actual signal values comprising the steps of: periodically sampling a received signal for producing a plurality of actual signal values;   predicting an actual signal value based on the plurality of actual signal values;   comparing the predicted signal value to the actual signal value, producing a delta value;   determining a maximum delta value for a group of N consecutive samples, where N is a positive integer; and   detecting an error if and only if the produced delta value is greater than the predetermined maximum delta value.   
     
     
       14. The method of claim 13 where N is greater than one. 
     
     
       15. A method of detecting and concealing errors in actual signal values comprising the steps of: predicting an actual signal value based on at least one previous actual signal value;   comparing the predicted signal value to the corresponding actual signal value, producing a delta value;   comparing the delta value to a predetermined maximum delta value;   detecting an error if and only if the delta value is greater than the predetermined maximum delta value; and   if an error is detected, concealing the error by replacing the actual signal value with the corresponding predicted signal value.   
     
     
       16. The method of claim 15 wherein the actual signal values are represented by periodic samples, the predetermined maximum delta value being determined for a group of N consecutive periods, where N is a positive integer. 
     
     
       17. The method of claim 16 where N is greater than one. 
     
     
       18. A method of generating error detection information from periodically sampled actual signal values, said method comprising the steps of: predicting an actual signal value based on at least one previous actual signal value;   comparing the predicted signal value to the corresponding actual signal value, producing a delta value; and   repeating said steps of predicting and comparing for a total of N actual signal values, where N is a positive integer, while: delaying the actual signal value a period of time corresponding to N sampling rates, and   storing the largest delta value as the error detection information.